JP3951370B2 - Plant operating condition monitoring method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a plant operating state monitoring method, and more particularly to a technique for monitoring a plant operating state using an artificial intelligence (AI) technique.
[0002]
[Prior art]
Conventionally, the following are known as techniques for monitoring the operating state of a plant.
(1) A specific upper and lower threshold value is set for measurement data measured for various devices constituting the plant, and the plant operating state becomes abnormal when the measurement data exceeds the upper and lower threshold value. An alarm is given to indicate that there is.
(2) Monitor fluctuations in measured data under a predetermined decision logic using an AI method such as an expert system.
[0003]
[Problems to be solved by the invention]
However, in the technique (1), it is not possible to perform detailed monitoring according to various operating conditions of the plant. In the technique (2), it is difficult to acquire knowledge and generate determination logic.
Moreover, although monitoring regarding the process amount is possible, it is difficult to monitor even the operation (operation data) performed by the operator on the plant.
[0004]
The present invention has been made in view of the above-described problems, and has the following objects.
(1) Provided is a plant operating state monitoring method capable of monitoring the plant operating state more finely.
(2) Provided is a plant operation state monitoring method capable of monitoring operation operation data by an operator.
(3) To provide a plant operating state monitoring method that facilitates acquisition of knowledge necessary for monitoring the plant operating state.
[0005]
[Means for Solving the Problems]
In order to achieve the above object, the present invention employs a means for determining the current operation state by processing one or a plurality of plant data related to the operation of the plant using a case-based reasoning technique.
In the above means, the processing using the case-based reasoning method is as follows:
a. A process of mapping the absolute value of each plant data sampled at regular intervals into a metric space;
b. Sequentially storing mapping information obtained by the step a as time-series data;
d. Based on the information obtained by the process a, a process of calculating the distance between the past time series data and the current time series data for each plant data;
e. A process of detecting past time-series data having a small sum of distances calculated in the process d as a similar operation example;
f. A method is adopted that includes a process of determining whether the current operation state of the plant is good or not based on similar operation examples.
Further, in the above means, g. The process of calculating the rate of change of each plant data;
h. A process of mapping the rate of change to a metric space;
i. A process of associating the rate of change with the absolute value based on time data associated with each plant data;
j. In place of the step b, the step of sequentially storing the mapping information obtained by the steps a and h and the association information obtained by the step i as time series data;
Adopt a means to add.
In addition, when at least one absolute value or change rate that is a distance greater than a certain value is detected in each of the above means, the past time series data is excluded from similar operation case candidates, and the past time series data The method of omitting the distance calculation of the other absolute value or change rate related to is adopted.
Further, in each of the above means, as the plant data, either one or both of measurement data measured for various process devices constituting the plant and operation operation data operated by the operator on the plant is adopted. Is adopted.
[0006]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an embodiment of a plant operating state monitoring method according to the present invention will be described with reference to the flowchart shown in FIG. In this embodiment, as will be described below, a case-based reasoning method, which is a kind of artificial intelligence (AI) method, is applied to monitoring the operating state of a plant.
[0007]
As a general form, a plant includes process equipment that performs a certain process, control equipment that controls the process equipment, and operation equipment that allows an operator to give the process equipment some operation. In the present embodiment, as data (plant data) related to the operation of the plant, measurement data that is variously measured for the process device and operation operation data that is input to the operation device by the operator are handled. In this case, each of the measurement data and the driving operation data is data for a single physical quantity or a collection of data for a plurality of physical quantities.
[0008]
First, such plant data is sampled at regular intervals and taken out for each physical quantity (step S1). There are various physical quantities in this case depending on the nature of the plant, such as flow rate, pressure, and electric power. In step S2, the absolute values of the plant data extracted in this way are mapped to the metric space. As shown in FIG. 2, for example, the absolute value is mapped to the metric space. Plant data related to steam flow (steam flow) is linked to each link in a predetermined tree structure (Steam Flow.0 to Steam Flow.6). (Mapping process) to be assigned to Here, each link is obtained by dividing the maximum fluctuation range of the steam flow rate so that there are a certain number of nodes (in this case, 3) having the lower part thereof.
[0009]
For example, in FIG. 2, when 0.58 is newly sampled, the plant data is Steam Flow. Assigned to 6 links. That is, the absolute value of the plant data sampled sequentially with the passage of time is mapped to one of the links according to the value, and is mapped to the metric space. Such mapping of absolute values to the metric space is performed for plant data of all physical quantities.
[0010]
When the mapping of the absolute value of the plant data to the metric space is completed in this way, a label indicating the characteristics of the plant data is assigned to each plant data (step S3). For example, in the case of plant data related to the above steam flow, labels such as “large steam flow”, “slight steam flow”, “normal steam flow”,... To grant.
[0011]
On the other hand, in step S4, the rate of change is calculated by comparing the previously sampled plant data with the current plant data. In a plant, not only the absolute value of plant data but also its tendency to change is often an important indicator. Therefore, normalize the plant data with its maximum value, and change by taking the difference (including the change direction) between this normalized plant data (normalized plant data) and the normalized plant data of the previously sampled plant data. A rate is calculated. The change rate is calculated for all physical quantities.
[0012]
When the change rate of the plant data is calculated in this way, the change rate is mapped to the metric space (step S5). The mapping of the absolute value to the metric space is performed in the same manner as the absolute value described above, and the rate of change sequentially calculated with the passage of time is set for each of the tree structures predetermined for the rate of change. It is a process assigned to a link.
[0013]
Then, when mapping of the change rate of the plant data to the metric space is completed, labeling is performed (step S6). For example, in the case of the above steam flow, depending on the rate of change of the steam flow, labels such as “high rate of change of steam flow”, “slightly high rate of change of steam flow”, “normal rate of change of steam flow”, etc. Is given for the rate of change of each physical quantity.
[0014]
As described above, the absolute value and the change rate of the plant data are independently mapped to the metric space and labeled. In step S7, the mapping information of the absolute value to the metric space and the change rate to the metric space are converted. Mapping information correlation processing is performed (step S7). In this correlation process, the rate of change and the absolute value are related based on time data associated with sequentially sampled plant data.
[0015]
When the labeling of the absolute value and the change rate of the plant data and the correlation process are completed in this way, the series of processing results, that is, the mapping information and the labeling information of the absolute value and the change rate of the plant data to the metric space. The related information is stored in the case memory as time series data (step S8). That is, time-series data is sequentially generated based on plant data sampled sequentially with the passage of time from the start of plant operation, and is sequentially stored in the case memory as operation examples.
[0016]
And with respect to the time series data (current time series data) of the newly sampled plant data, one or a plurality of time series data (past time series data) over the past fixed time width already stored in the case memory Are compared and past time series data (past driving case) similar to the current time series data (current driving case) is retrieved as a similar driving case (step S9).
[0017]
Specifically, the mutual distance between the data is calculated for all combinations of one or more past time-series data and current time-series data over a certain past time span. That is, the distance between the past time-series data and the current time-series data is calculated for all absolute values and change rates mapped to each link by the mapping process to the metric space (steps S2 and S5). Then, the past time series data having the smallest sum of the distances is detected as a similar operation case most similar to the current time series data.
[0018]
For example, in FIG. 2, the absolute value of the plant data belonging to the same link has the smallest distance. And the distance which is in the relation which reaches | attains following more links becomes large. Steam Flow. The absolute value of 0.34 belonging to link 4 is Steam Flow. The absolute value 0.44 belonging to the link 5 is Steam Flow. 2 links can be followed.
[0019]
In contrast, Steam Flow. The absolute value of 0.34 belonging to link 4 is Steam Flow. The absolute value of 0.68 belonging to the link of 3 is Steam Flow. It can be traced via a zero link. Therefore, the absolute value 0.44 has a smaller distance than the absolute value 0.68 with respect to the absolute value 0.34 because the number of links that need to be traced is small.
[0020]
Such distance calculation of each plant data is performed for all combinations of one or more past time-series data and current time-series data over a certain past time width. However, in order to shorten the processing time, if at least one absolute value or change rate that is a distance greater than a certain value is detected, the past time series data is excluded from similar operation case candidates, It is conceivable to omit distance calculation of other absolute values or change rates related to past time series data.
[0021]
When a similar operation case is detected in this way, the current operation state of the plant is determined based on the similar operation case, and a message is output (step S10). In other words, since this similar operation case shows the current time-series data, that is, the operation state of the plant similar to the current operation case, if the similar operation case is determined to be normal operation in the past, any message is output. do not do.
[0022]
On the other hand, if the similar operation case is determined to be abnormal operation, a message notifying the abnormal state is output assuming that the current operation case is in an abnormal operation state. If a similar driving case is not detected, the current driving state is an inexperienced driving state in the past, so a message indicating that fact is output. By outputting the message in this manner, it is possible to alert the operator about the operating state of the plant.
[0023]
In the above embodiment, the number of links that need to be traced for distance calculation is used. However, the present invention is not limited to this, and the absolute value of the difference between two numerical values may be used more easily by taking advantage of the characteristics of numerical data. . In the above embodiment, measurement data and operation data are taken as plant data. However, the present invention is not limited to this, and either one may be handled. The time width set for past time series data for comparison with the current time series data and which physical quantity of plant data to compare with respect to past time series data and current time series data are monitored. It is necessary to set as appropriate based on the characteristics of the plant.
[0024]
【The invention's effect】
As described above, the plant operation state monitoring method according to the present invention has the following effects.
(1) By using the case-based reasoning method for monitoring the operation state of the plant, the operation state of the plant can be monitored more finely than the conventional monitoring using a constant threshold value.
(2) By using the operation data as the plant data input to the case-based reasoning, it is possible to monitor the operation of the operator for the plant that has not been performed so far.
(3) By using case-based reasoning, operation cases are automatically accumulated in parallel with the operation of the plant, so that knowledge necessary for monitoring the operation state of the plant can be easily obtained.
(4) The operation state of the plant can be monitored more accurately by applying the case-based reasoning not only to the absolute value of the plant data but also to the rate of change thereof.
[Brief description of the drawings]
FIG. 1 is a flowchart showing an embodiment of a plant operation state monitoring method according to the present invention.
FIG. 2 is an explanatory diagram for explaining a process of mapping plant data to a metric space in one embodiment of a plant operation state monitoring method according to the present invention.

Claims (6)

By processing one or more plant data related to plant operation using a case-based reasoning method, the current operating state is determined,
The process using the case-based reasoning method is as follows:
a. The process of mapping the absolute value of each plant data sampled at regular time intervals into a metric space by mapping to a predetermined tree structure ;
b. Sequentially storing mapping information obtained by the step a as time-series data;
d. Based on the mapping information obtained by the step a, a step of calculating a distance between past time series data and current time series data for each plant data;
e. A process of detecting past time-series data having a small sum of distances calculated in the process d as a similar operation example;
f. And a step of determining whether the current operation state of the plant is acceptable based on the similar operation example. In the plant operating state monitoring method according to claim 1,
g. The process of calculating the rate of change of each plant data;
h. A process of mapping the rate of change to a metric space;
i. A process of associating the rate of change with the absolute value based on time data associated with each plant data;
j. In place of the step b, the step of sequentially storing the mapping information obtained by the steps a and h and the association information obtained by the step i as time series data;
A plant operating state monitoring method characterized by comprising: The plant operation state monitoring method according to claim 1 or 2, wherein
When at least one absolute value or change rate that is a distance greater than a fixed value is detected, the past time series data is excluded from similar operation case candidates, and another absolute value or change rate relating to the past time series data is excluded. The plant operating state monitoring method characterized by omitting the distance calculation of The plant operation state monitoring method according to claim 1, wherein the plant data is measurement data measured for various process devices constituting the plant. The plant operation state monitoring method according to any one of claims 1 to 3, wherein the plant data is operation operation data that an operator operates on the plant. The plant operation state monitoring method according to any one of claims 1 to 3, wherein the plant data is measurement data measured for various process devices constituting the plant and operation operation data operated by the operator on the plant. A method for monitoring the operating state of a plant.

Images